Monday, March 7, 2016

Moderation and comment policy

Do not offer sympathy to the mentally ill. Tell them firmly, "I am not paid to listen to this drivel. You are a terminal fool."
‒William S. Burroughs
Look. I really like to argue about technical subjects. I like to be told I'm wrong because of X. That's because even if I don't think I'm wrong and argue back that X is incorrect at least I either now know that X exists or that "you're wrong because of X" is an argument I might encounter in the future. Like most humans, I make mistakes from time to time and sometimes X points out that mistake. I've actually taken down a post (placeholder is here) because a commenter (Tom Brown) pointed out an X and it made me realize I had approached the whole thing incorrectly.

But let me provide a list of helpful tips for arguing with me in comments:

• Please do not leave off X. If you do, I will assume that I've just made you mad and you don't actually know what your problem is. The comment will probably be deleted unless I can think of a funny response.
• If X is about physics or math, be aware that I have a PhD in physics and BS in mathematics (and many years working as an R&D scientist in industry). Like all humans, I will occasionally (frequently?) make a sign error (like here, again thanks to Tom Brown) or drop a factor of 2. However you should realize the prior probability that my expertise in these areas is greater than yours is almost 1. Don't tell me what you think of curve fitting, Taylor expansions, or effective field theory. Unless you say something extremely insightful or provide details, I will probably respond with snark.
• If X is about something you yourself have never studied in school, please note that Dunning-Kruger effect is a hell of a drug. It's also my favorite thing to point out in responses to smarmy but ignorant comments. While sufferers of the effect are probably the least capable of determining their condition, at least give it a try.

Feel free to bug me on Twitter using your egg avatar, though.

*  *  *

Original post (before 1 March 2017 update):

It is with some reluctance that I've decided to enable comment moderation due to abuse, ad hominem attacks, and terminal fools.

Since this is my blog, I can do what I want. And since I'm an elitist bastard, I'm going to go a bit further than others do and ban stupid comments, failures of reading comprehension, and the rather baffling inability to Google something before you post (you're on the Internet already).

Additionally, all pro-"Post Keynesian" comments are also banned. This school of thought apparently consists of terminal fools. The determination of Post Keynesianism will be made via a very sophisticated machine learning algorithm so it might be some time before sufficient data has been accumulated. Please stand by.

Furthermore, out of spite, I am now a Market Monetarist. While they are also wrong, they are a much better class of people.

They also don't think economics is "just accounting".

...

Update 12 March 2016: I think I may just turn moderation off and on as it becomes necessary e.g. because I say something that is correct but inflammatory towards a particular viewpoint.

Update 1 March 2017: I have turned off anonymous commenting. It was being abused by a knowledge-impervious dolt suffering from Dunning-Kruger. In general, the quality of comments from anonymous commenters has declined significantly over the past few years. Requiring Open ID at least lets me figure out which anonymous commenters are which and block them accordingly. I also rewrote the opening and will place a link here on the side bar.

1. Jason,

I am probably one of the terminal fools you are referring to so I'll take my leave and wish you the best of luck with your project.

Your quote about terminal fools from Burroughs is a little ironic given his grandfather was in the business of manufacturing terminals, but perhaps not of the lamebrain variety.

Henry.

1. Henry,

You at least have the curiosity -- even when we've had a long back and forth, I still feel like it is possible we could see eye to eye. There is usually engagement with the issue at hand.

2. Furthermore, out of spite, I am now a Market Monetarist. While they are also wrong, they are a much better class of people.

Been there, done that.

I can't say that I blame you. ;-)

3. BTW Jason the above was not my comment.

"You at least have the curiosity"

I was genuinely interested in what you are doing.

Anyway, all the best.

Henry.

2. Hey Jason

Im sorry you felt compelled to write this post. I know you have only been trying to give an honest critique to ALL economic schools of thought and you have a very informed and unique perspective, you aren't just shooting form your hip. I do have to come out in defense of Pkers however and Ill give you my reasons in a bit.

Erase this post if you must but Im only offering my own perspective in the same vein I believe you have offered yours, as a practitioner in another field (medicine, anesthesia particularly) that does require a level of analysis and perspective that at the very least allows me to see through a lot of stuff that gets presented in our economics literature. I didn't come to my exploration of finance/econ from any particular school and am healthily (I think) skeptical of common wisdom. This is probably why heterodox thinking is attractive to me.

Now your IQ is probably 3x mine so Im not trying match intellect and dig specifically into any of your posts where you did your criticism of SFC models. I simply want to defend the what is dismissed as the "just accounting" approach.

For me, its clear that accounting is simply the way we measure what has happened. Depending on what your looking at can determine what accounting methods you use but we MUST have accounting in order to do economics in a society using money and banking. And one thing about accounting is it only tells you what already happened not what will happen. It can tell you what CAN happen though. If you account back far enough you can tell a story about how the accounting got to where it is. Theoretically we can have an accounting of everyones accounts over each 24 hr period of the last 25 years (longer actually) and having this information would be like looking at the fossil record to explain biological evolution. The more details you know about the accounting record the more accurate a story you can tell. Trying to do economics and prescribe policy without a firm appreciation and understanding of accounting sounds like malpractice to me.

PK adds way more to the body of economic knowledge than monetarism. Monetarism is simply a religion with high priests who can never be dissuaded. There isn't a way to disprove their beliefs, just as you cant disprove god.

Anyway, I look forward to seeing where you take your thinking on this site.

Thanks

1. Hi Greg,

The market monetarist comment in my post above was a joke. Additionally, I don't usually have a lot of problems with PK policy prescriptions or political goals.

But the "just accounting" approach has some fundamental issues. I've just come up with a new way to look at it; let me try it out on you.

Accounting is essentially:

cash in - cash out + opening balance = closing balance

or

C1 - C0 + B0 = B1

There are two ambiguities (invariances) in this equation. One is that I can multiply everything by a constant (X → α X) because:

α C1 - α C0 + α B0 = α B1

C1 - C0 + B0 = B1

That is a scale invariance, and it is common to all of economics, because the unit of currency isn't pinned down. However, there is a second invariance (ambiguity) ... that of F → F + ϕ and S → S + σ (where F is a flow and S is a stock) so that:

C1 + ϕ - (C0 + ϕ) + B0 + σ = B1 + σ

C1 - C0 + B0 = B1

This invariance is problematic for accounting because it means that it is invariant to levels of stocks and flows -- and not just certain combinations, but independent changes of the levels of stocks and the levels of flows.

This would be fine if all levels of stocks and flows were linear in time. Say your balance goes from 1 to 2 to 3:

1 - 0 + 2 = 3
1 - 0 + 1 = 2
1 - 0 + 0 = 1

The growth rate is 1 unit per unit time. Let's try our invariance by adding 10 to our balance:

1 - 0 + 12 = 13
1 - 0 + 11 = 12
1 - 0 + 10 = 11

The growth rate is the same: one unit per unit time.

However, a pure accounting approach is problematic because accounting can't specify changing rates of change between time periods where the accounting holds.

Let's say your balance goes from 1 to 2 to 4 like this:

2 - 0 + 2 = 4
1 - 0 + 1 = 2
1 - 0 + 0 = 1

That's a continuously compounded growth rate of 69% per unit time. But remember, the accounting equation is invariant under independent shifts of stocks and flows. Let's add 10 to the stocks:

2 - 0 + 12 = 14
1 - 0 + 11 = 12
1 - 0 + 10 = 11

This situation has a growth rate that is changing in time (it increases from 12% to 20% from the first to the second time period). That is to say: the growth rate isn't invariant under the same transformations as accounting. Basically, accounting doesn't pin down growth rates.

This was the problem in the previous post: the accounting equations didn't pin down the rate of growth between time periods leaving a degree of freedom (in the example above, it is the growth rate).

Since not only growth but changing growth is a major feature of macroeconomics, an "accounting only" approach will be ambiguous.

What the stock-flow approach tries to do is extract dynamics from accounting. This only works if there is linear change. If the change coming out of an SFC model appears to be non-linear, then there must be dynamical assumptions buried somewhere.

Now these dynamical assumptions are fine -- but they aren't "just accounting".

2. Interesting take.

3. Jason, wouldn't these problems affect other discrete time models, for instance the New Keynesian model that Nick Rowe gets free mathematical consulting services with on his blog here.

Also, for SIM, it's very clear where alpha1, alpha2 and theta contribute to the dynamics. They aren't touted to be "just accounting," are they? They explicitly describe assumptions about the behavior of agents and the government. Now I realize there are still some problems there, but I never thought for a minute that the dynamics of the system were independent of these behavioral assumptions.

4. It wouldn't affect all models because not all finite difference models have this same invariance. NK DSGE models are also written in log linear variables which don't have the same issues.

The alphas (along with the various identities) are just a linear transformation -- overal they have little effect on the dynamics. You could imagine finding a solution for just two variables (H and G-T) and the perform a linear transformation to obtain any other solution.

5. "You could imagine finding a solution for just two variables (H and G-T) and the perform a linear transformation to obtain any other solution."

I'm sure you could... but when I first looked at it I saw H as the one state variable and its time constant (Tc) completely determined as such:

Tc = -Ts/log(1 - θ*α2/(1 - α1 + θ*α1)) = 5.9861 periods

Where Ts=1 is the initial sample period. I.e only a function of Ts, α1, α2 and θ. There's an issue with how they try to scale this to a new sample period (Ts2): essentially by applying a time scaling constant c = Ts2/Ts like this:

Tc2 = -Ts*c/log(1 - θ*α2*c/(1 - α1 + θ*α1))

The problem being that Tc2 ≠ Tc, but IMO that's because they're effectively approximating exp(at) ≈ 1+at.

If that approximation holds (i.e. |at| << 1), then

Tc2 ≈ -Ts*c/(-θ*α2*c/(1 - α1 + θ*α1))
= Ts*(1 - α1 + θ*α1)/(θ*α2) ≈ Tc

Substituting in their initial parameter values we have (1 - 0.6*(1-0.2))/(0.2*0.4) = 6.5 periods, so it's not a great approximation, but that would require |θ*α2/(1 - α1 + θ*α1)| << 1, and it's 0.154... perhaps not really small enough. If initially alpha2 = 0.01 it would be better (260 periods vs 259.5 periods).

But sticking with Ts=1, that time constant (Tc≈6 periods) holds no matter the initial value of H, or what the exogenous input G is. T is just an output in my formulation: a scaled version of the single state variable H offset by a scaled version of exogenous input G: so naturally its dynamics will be determined by those two.

6. Also, in your examples, if you were to measure the growth of case 3 as units per unit time (as you do in cases 1 and 2):

Case 3:
2 - 0 + 2 = 4
1 - 0 + 1 = 2
1 - 0 + 0 = 1

It would be 1 the 1st period and 2 then second. However, if you were to compare cases 1 and 2 on the bases of continuously compounded growth rates (as you do case 3, which comes out a constant 69%), then we'd have:

Case 1: (1 -> 2 -> 3)
1st period: 69%
2nd period: 41%

Case 2: (10 -> 11 -> 12)
1st period: 10%
2nd period: 9%

7. Tom, it's not appropriate to measure linear growth with a continuously compounded growth rate.

In the second one growth is changing (and in the simple 2^t model it eventually approaches log 2 growth rate) so the continuously compounded rate is the only one that makes sense.

8. Sure, I agree. It didn't occur to me until after I wrote that exactly what your point was (although your examples were clear enough). Missing that delete button! Lol

9. ... and to be more clear above, I should have written:

"...that's because they're effectively (implicitly?) approximating exp(-t/Tc) ≈ 1 - t/Tc."

3. Thanks for taking time to respond Jason

First, I don't know of any PK economists that I have read who suggest that our accounting systems are infallible and therefore their models are bulletproof. So I think many would welcome efforts to help us improve them in order to provide a better explanation of the past and to hopefully make better prescriptions for the future.

Im not sure I totally agree with this.....

"Accounting is essentially
cash in - cash out + opening balance = closing balance"

at least not that thats ALL accounting is. Accounting also involves understanding the forces that change opening and closing balances (which in many cases are asset prices) beyond just adding more cash to them. These involve understanding the dynamics of debt creation and how these can negatively and positively affect various prices. So the dynamics of that equation are different for a stock account and a checking account

I think one thing that is a particularly vexing issue is how to accommodate for what is typically described as different properties of money eg unit of account and unit of exchange. I see this as the largest area of confusion amongst the various econ schools

In your above simplification, depending on the account type, the opening balance can be a collection of bond and stock prices (like my 401k) plus some cash or just cash or no cash. It might also just be a passbook savings where I just add up my monthly cash contributions to get a gradually growing number. If its a stock account that is worth 100,000$I might add cash of 10,000$ (giving me nominally 110,000$overall) but a week later the market tanks 25% and I now have stocks worth 75,000$ and 10,000$cash for 85,000$ total. Have I "lost money"? Maybe this is why stocks and flows appear be not linear in time.

It seems to me that many people (not saying you necessarily) confuse and conflate the two things, thinking of a rising value of their 401 k as "having more money" (wealth affect?). I feel like NKers try to separate out things better by talking about asset prices, deposit balances, interest rates (both as earnings AND costs) instead of lumping them all together into "money".

A lot of the disagreement amongst the schools is language and I feel like NK are the most consistent with their language and try and separate out the various components of the financial system instead of just referring to it all as money. In fact most will not use the term money without describing exactly which part of money they mean. MMers aren't nearly so precise and careful.

Im not sure that StockFlows try and extract dynamics but I do think they don't just hand wave and explain rising stock prices as expectations and fully dependent on CB policy. They do try and give real explanations plus they fully explain operations like QE as asset swaps rather than the dreaded "money printing" that gets everyone aghast.

Anyway, Ive already gone on too long and I don't want to come across as trying to persuade you into being a NKer. I just want to give you the state of my obviously imperfect understanding of things. I do look forward to seeing more of your posts

btw Im reading book by Cesar Hidalgo that is in the same vein as this site, interesting stuff

1. Greg, you repeatedly write NKer above, but do you mean PKer?

2. Hi Greg,

You said:

At least not that thats ALL accounting is ...

That may be true, but the issue I discussed above and originally pointed out as a problem in Godley and Lavoie was in the accounting matrix where columns and rows sum to zero. And that matrix was the source of the dynamics in the model. That is to say the indeterminacy is embedded in my definition of accounting above. If accounting is more than my definition, it still doesn't clear up the problem with accounting matrix or SFC models.

You said:

It seems to me that many people (not saying you necessarily) confuse and conflate the two things, thinking of a rising value of their 401 k as "having more money" (wealth affect?). I feel like NKers try to separate out things better by talking about asset prices, deposit balances, interest rates (both as earnings AND costs) instead of lumping them all together into "money".

You might be interested in my take on this -- different things are "money" depending on circumstances:

http://informationtransfereconomics.blogspot.com/2015/07/kaldor-endogenous-money-and-information.html

Sometimes stocks are liquid assets comparable to cash (i.e. are accepted much like cash); sometimes not.

You said:

stock prices as expectations and fully dependent on CB policy ...

Here's another link from me on this:

http://informationtransfereconomics.blogspot.com/2015/04/solving-dark-matter-problem.html

In the information transfer view things can be more than the sum of their parts. And this stock value only exists in the aggregate -- it may be impossible to determine from accounting. In the economy there may be an entropy term that would be missing if you just tried to add up all the individual products. This entropy term can be thought of as an entropic force acting against agents undoing their economic gains.

4. Thanks Tom, I did mean PKer.

Jason, you are certainly correct, as I see it anyway, that the moneyness of things is not fixed. Stocks for example can be very liquid, certain ones anyway but they will never reach cash on the liquidity scale. Certain stocks may pass certain bonds at various times as might certain hard commodities like silver. There is room for some fluctuations in the non cash levels of the hierarchy to be sure. And CB policies can certainly have affects on these.

" things can be more than the sum of their parts. And this stock value only exists in the aggregate -- it may be impossible to determine from accounting."

Certainly the object of proper accounting isn't to determine a stock price so Im not sure what you are getting at here. But if a stock value is being used on a balance sheet to determine things like solvency or ability borrow/lend there are less or more proper ways to account for them.

For me, appreciating the dual nature financial instruments.... that they have an asset and liability side has allowed me to think more clearly about things like govt debt, stock market values, private lending, QE and other CB policies and even notions like saving, investment, spending and consuming/producing. Its really quite shocking to me to see people who are paid to comment on policy and are supposedly experts make some of the statements they do about our financial system...... they are f*3k!ng clueless! Its all well and good that we debate certain things but if you don't understand that we are no longer on a gold standard (thank god) and how that changes the narrative of finance you should just sit down and shut up about these matters. I have almost as strong a views about the "Quantity theory of money" as well. Its not as inapplicable but it should never be used in the strong form, in my view.

I definitely like where you are looking with regards to the entropy term. No doubt there is something missing form our present models.

5. Re: re-updated comment policy (modulated moderation?). I like it!